Fit a regression per group on one crossplot

For DG3 poroperm transforms, you typically want one regression line per facies (or per zone, or per well) on a single crossplot — each line in its own colour, with its R² and equation in the legend, all in one figure. logSuite does this in two lines.

The minimal pattern

from logsuite import Crossplot

xplot = Crossplot(manager, x="PHIE", y="PERM", color="Facies",
                  equation_format="petrel", decimals=3)
xplot.add_regression_per("Facies", "exponential", legend_loc="upper left")
xplot.save("poroperm.png")

Three concrete things this gets right without any extra work:

  • Line colours match the manager palette — set manager.Facies.colors = {0: "#999999", 1: "#3b82f6", 2: "#10b981"} once and every regression line picks up the matching colour.

  • Legend equations match what Petrel expects — because equation_format="petrel" was set at the constructor, every fit’s legend shows pow(10, c1*x + c0).

  • No monkey-patching of xplot._data — add_regression_per enumerates the unique facies values internally and calls add_regression(where={"Facies": [value]}) for each.

When you need an arbitrary subset

For a custom mask (e.g. the high-PHIE half of one facies), use add_regression(where=callable) and get the internal column names from column_for:

facies_col = xplot.column_for("Facies")
phi_col = xplot.column_for("PHIE")

xplot.add_regression(
    "exponential",
    name="Clean (PHIE>0.20)",
    where=lambda df: (df[facies_col] == 2.0) & (df[phi_col] > 0.20),
    line_color="#fbbf24",
    line_style="--",
)

Subsets smaller than min_samples (default 5) emit a warning and are skipped — no exception.

Verifying

Two runnable examples in story_tests/:

  • story_1_regression_subset.py — per-facies fits with palette colours and Petrel-form equations.

  • story_5_per_group_regressions.py — same plus an arbitrary callable subset.